# FluxForge

A configuration-driven fine-tuning framework for Hugging Face Diffusers pipelines, such as Stable Diffusion, where a YAML file describes the whole training pipeline, including which layers to freeze or train, instead of a hand-written training script.

FluxForge is a fine-tuning framework for diffusion pipelines built on Hugging Face Diffusers. Instead of writing an imperative training script, a user describes the whole run, including data loading, the optimizer, the learning-rate scheduler and checkpoint serialization, in one YAML configuration file that can be versioned and shared.

It supports modular, layer-frozen training of the diffusion UNet or transformer backbone, so individual components, such as cross-attention layers, can be frozen, partially trained or fully fine-tuned while others stay fixed.

It requires Diffusers 0.30 or later, PyTorch 2.1 or later and Python 3.10 or later, and is licensed under MIT.

Website: https://github.com/njgymb/diffusers-sculptor